node 18.13 potential memory leak

[francois-gsk]

Version

18.13

Platform

linux/amd64

Subsystem

docker linux/amd64 alpine 3.17

What steps will reproduce the bug?

I only upgrade from 18.2 to 18.3

How often does it reproduce? Is there a required condition?

No specific condition

What is the expected behavior?

when I rolled back to node:18.12:

What do you see instead?

Memory stats since the upgrade to node:18.13:

Additional information

Code base very huge, not possible to exactly know where is the leak, we rolled back ASAP because our prod env was crashing.

I posted also an issue on the node docker github repo.

[marco-ippolito]

I think it's impossible to find where the memory is leaking without an example of what you are doing in your code

[RoyMusthang]

it would be good if you gave the code you used to do this test, without it there is no way to replicate and locate the error

[ephys]

We'll try to produce a minimum reproducible example as soon as we can :)

[jamesdiacono]

Could this be related to #46435?

[francois-gsk]

Could this be related to #46435?

Yes, one of the (few) task performed on a minute basis is a fetch.

[ephys]

I agree, it's very likely that the culprit was fetch. Most elements in my heap snapshots that are not going away are fetch related, and the issue seems to only be present in the couple of endpoints that use undici's fetch

[sirpy]

I have similar issue, didnt verify it is from >18.12 but when i upgraded to node 18 from node 16
i've started seeing memory leaks on our heroku node.
when i've upgraded to node 20 or downgraded to node 16 issue was solved